Chocolate tracks with Nobel prizes, and other lessons in telling a link from a cause

Key takeaways

• A 2012 paper found that countries eating more chocolate had more Nobel laureates, and its author called the result a hypothesis; a study within a large cocoa trial later found no benefit for thinking skills.
• In a study of more than 400,000 older adults, coffee drinkers first looked more likely to die, until researchers accounted for the fact that they smoked more.
• A review written by supplement-industry experts found that supplement users are better educated, exercise more and smoke less, which can make pills look better than they are.
• After hundreds of studies linked low vitamin D to many diseases, researchers proposed that illness may lower vitamin D rather than the other way around.
• Before you change a habit because of a headline, ask whether a hidden factor, a reversed arrow or plain luck could explain the link.

In October 2012, the New England Journal of Medicine, one of the most respected medical journals in the world, published an unusual graph. Its author, Franz Messerli, plotted chocolate eaten per person in 23 countries against Nobel Prizes per 10 million people¹.

The fit was tight. According to a summary on NEJM Journal Watch’s CardioExchange blog, the correlation was 0.79, on a scale where 1 is a perfect match¹. Switzerland led in both chocolate and laureates¹. Sweden was the outlier: by Messerli’s line it should have had 14 laureates, but it had 32²³.

Messerli disclosed that he eats chocolate every day, mostly Lindt’s dark varieties¹. He also wrote that his finding could only generate hypotheses, to be tested in a randomized trial¹.

Some coverage skipped that caution. A food-industry news site ran the headline “Eating chocolate produces Nobel prize winners, says study”³.

That leap, from “goes with” to “produces,” happens in health news every week. The good news is that most false leaps fall into three predictable traps. Once you know them, you can test a headline in about a minute.

A correlation says two things move together, and that’s all

A correlation measures how closely two things rise and fall together. It runs from minus 1 to plus 1. Zero means no pattern, and the closer to 1 (or minus 1), the tighter the pattern.

What a correlation can’t tell you is why. A pattern between A and B can appear for at least four reasons besides A causing B:

• A hidden third factor drives both. Scientists call this confounding.
• The arrow points backward. B causes A, which is called reverse causation.
• Chance. Random noise sometimes lines up by accident.
• Many looks. Test enough pairs and some will line up by luck, a problem called multiple comparisons.

The chocolate graph invites all of them. Messerli himself raised the reverse possibility, that sharper minds might seek out more chocolate¹. One trade report noted that the study left out the economic factors that give some nations easier access to chocolate³. Wealth also buys universities and research labs, which makes it a plausible hidden factor, though that is our reasoning, not a tested finding.

Suspect one: a hidden third factor

Coffee shows how a hidden factor can flip a result.

In 2012, Freedman and colleagues reported in the New England Journal of Medicine on the NIH-AARP Diet and Health Study. As a published commentary on the paper summarizes it, the study followed 229,119 men and 173,141 women aged 50 to 71 from 1995 to 2008⁴. When the researchers adjusted only for age, coffee drinkers were more likely to die during the study⁴.

The coffee drinkers, it turned out, smoked more often⁴. Once the researchers accounted for smoking and other factors, the picture reversed. Men drinking four to five cups a day had a 12% lower risk of death, and women 16% lower⁴.

The authors were careful. They wrote that their data could not show whether the link was causal⁴. A factor they didn’t measure could still be shaping the result.

The same hidden factor shaped coffee’s cancer reputation. In 1991, the International Agency for Research on Cancer (IARC), part of the World Health Organization, classified coffee as possibly carcinogenic⁵. In 2016, a panel of 23 scientists reviewed more than 1,000 studies and moved coffee to Group 3, meaning not classifiable⁵.

The panel found no cancer-causing effect for pancreatic, female breast or prostate cancer, and lower risk for liver and uterine cancers⁵. IARC said the earlier bladder cancer links may have come from other risk factors, such as tobacco smoking⁶. Its listings show how strong the evidence is, not how big a risk is⁶.

There’s a twist worth knowing. The same panel classified very hot drinks, above 65°C (149°F), as probably carcinogenic⁵. The temperature, not the coffee, drew the concern.

Supplement users are different people

Confounding matters most when a product appeals to people who already live healthier lives. Supplements are a textbook case, and some telling evidence comes from the industry itself.

In 2014, two experts tied to the Council for Responsible Nutrition, the supplement industry’s trade group, reviewed research on who takes supplements⁷. They found users tend to be better educated, earn somewhat more, exercise more, avoid tobacco and eat slightly better diets⁷.

The gaps are large. In federal survey data from 2003 to 2006, supplement use reached 61% among adults who went beyond high school, versus 37% among those who never finished it⁷. In the 1999 to 2000 survey, former smokers used supplements far more often than people still smoking, 61% versus 43%⁷. The trade group’s foundation paid for the review, which makes its findings about users’ advantages hard to dismiss as hostile⁷.

Now imagine a study that compares supplement users with nonusers. Every one of those advantages could make the pills look protective even if they did nothing.

A 2019 study in the Annals of Internal Medicine, led by researchers at Tufts University, tested this with 30,899 U.S. adults followed for a median of 6.1 years⁸¹⁴. Supplement use as such was not linked to a lower death rate⁸. Getting enough vitamin A, vitamin K, magnesium, zinc and copper was linked to lower death rates from all causes or from heart disease, but only when those nutrients came from food⁸.

The same study found that 1,000 milligrams or more a day of supplemental calcium was linked to higher cancer death rates⁸. Apply the same rule here: that’s a link too, and the authors noted their results could reflect leftover confounding⁸. If you take calcium for a medical reason, talk with your clinician before changing anything.

For readers who value natural living, there’s encouraging news here. Whole food, regular movement and not smoking tend to travel together, and those habits show up again and again in the data. The hard part is crediting any single product for what a whole way of living may be doing.

Suspect two: the arrow points the other way

Sometimes the effect causes the supposed cause. Vitamin D is a well-studied example.

In 2014, a team led by Philippe Autier reviewed 290 prospective cohort studies and 172 randomized trials of vitamin D in The Lancet Diabetes & Endocrinology⁹. Most cohort studies linked low blood levels of vitamin D with heart disease, infections, mood disorders, cognitive decline, death and more⁹.

The trials told another story. Giving vitamin D did not prevent those diseases⁹. The one hint of benefit was a slight drop in deaths among older adults, mostly women, taking 20 micrograms a day⁹.

Why the gap? The authors proposed that inflammation from illness lowers vitamin D levels, so low vitamin D may be a sign of poor health rather than its cause⁹. They noted that this remains unproven⁹.

A later trial added nuance. In VITAL, which randomly assigned 25,871 older adults, people taking 2,000 IU of vitamin D3 a day had a borderline 22% lower rate of confirmed autoimmune disease than those on placebo over a median of 5.3 years, a 2022 BMJ report found¹⁵. Vitamin D still matters for bones, and a diagnosed deficiency is a separate question to discuss with your clinician.

Reverse causation hides in everyday claims, too. Picture a study finding that people who take sleep aids report worse sleep. The likeliest story is that poor sleepers reach for sleep aids, not that the pills ruin sleep. (This is a hypothetical example.)

Suspect three: luck, especially when researchers look many times

Chance is easy to underestimate. In the late 1980s, a trial called ISIS-2 tested aspirin in 17,187 people with a suspected heart attack¹⁰. Aspirin’s overall benefit was overwhelming, with odds of a chance result below 1 in 100,000¹⁰.

When The Lancet’s editors asked for results broken down by patient subgroup, the investigators agreed on one condition: the table would also split patients by astrological sign¹¹. For Gemini and Libra, aspirin appeared slightly harmful, a difference small enough to be noise¹⁰. The point was a warning against trusting any single subgroup finding¹¹.

Peter Sleight, an Oxford researcher, later spelled out the arithmetic. Split a trial with no real effect into 40 subgroups, and you should expect about two “significant” results by luck alone¹⁰.

Nutrition studies face the same math on a bigger scale. A cohort that asks about dozens of foods and tracks dozens of diseases creates hundreds of possible pairs. A few will look impressive by accident. A single surprising link, never repeated, deserves a shrug.

How scientists judge cause: Bradford Hill’s nine viewpoints

In 1965, the British researcher Austin Bradford Hill gave a talk to the Royal Society of Medicine’s occupational medicine section. Published that year, it offered nine viewpoints for deciding when a link is likely causal¹². They are still widely cited, often under the name “criteria”¹³.

Hill insisted these weren’t hard rules, and that a cause can be real even when some viewpoints don’t apply¹². He also warned against putting too much faith in tests of statistical significance¹². Jeffrey Aronson, writing on The BMJ’s opinion blog in 2018, argued that only one viewpoint, timing, works as a true criterion: if the supposed cause came after the effect, that strongly argues against it¹³.

Here are the nine, rewritten as questions and tested on the chocolate claim. The right-hand column is our own assessment.

Viewpoint	The question to ask	Chocolate and Nobel Prizes
Strength	How big is the link?	Strong at the country level1
Consistency	Do other people, places and times show it?	A similar country-level pattern turned up for milk2
Specificity	Is it tied to one exposure and one outcome?	No, if milk works just as well2
Timing	Did the exposure come first?	Unclear from country averages
Dose	Does more exposure bring more effect?	Implied by the trend line, but only for countries, not people
Plausibility	Is there a believable mechanism?	Cocoa flavanols were proposed2
Coherence	Does it fit what else is known?	Weak; prizes reflect decades of work and funding
Experiment	Does changing the exposure change the outcome?	A large trial found no cognitive benefit from cocoa extract2
Analogy	Have similar causes done similar things?	No clear parallel

That trial was COSMOS. Its COSMOS-Mind study, with 2,262 participants, found no benefit of cocoa extract for thinking skills, McGill University’s Office for Science and Society reports².

A headline decoder you can use today

The five headlines below are invented for teaching. Each is the kind you’ll see, paired with its likeliest trap.

Sample headline (made up)	Likeliest suspect	Question to ask	What would convince you
“Green tea drinkers have younger-looking skin”	Hidden factor	Did they account for sun protection, smoking and income?	A trial assigning tea or no tea
“Long sleepers die sooner”	Reversed arrow	Could illness cause the long sleep?	Results that hold after excluding people who were ill
“Organic shoppers have lower cancer rates”	Hidden factor	Do organic shoppers also smoke less and exercise more?	Similar results in groups with similar habits
“Turmeric eaters show 18% less of one cancer”	Luck and many looks	How many foods and cancers were tested?	The same link repeated in a separate study
“Nations eating more fermented food live longer”	Hidden factor, country data	Is this about countries or individuals?	Studies that follow individual people

Notice that none of these headlines is necessarily false. Each is unproven, and the decoder shows what proof would look like.

A one-minute routine for any health claim

1. Find the verb. “Linked to,” “associated with” and “tied to” signal a correlation. “Caused,” “cut” or “raised” should come from a trial.
2. Name the study type. Did people choose the habit, or did researchers assign it by chance? Our companion article on study designs explains the difference.
3. Run the three suspects. Who else might differ? Could the arrow point backward? How many things were tested?
4. Check the size. A 10% difference is fragile. A doubling or more is harder to explain away.
5. Look for an experiment. If a trial exists, it outranks the headline.
6. Weigh the stakes. A cheap, safe change like adding vegetables needs less proof than an expensive pill or stopping a prescribed treatment.

Your own experience deserves the same respect and the same questions. If you started a remedy and a cold faded, the timing is real, but colds fade on their own. Our article on healing stories looks at how to weigh personal experience.

What the evidence says about the examples

Claim	Evidence level	Basis
Coffee itself helps people live longer	Promising but early	A huge observational study found lower death rates after adjusting for smoking, but its authors could not show cause4
Coffee causes pancreatic, breast or prostate cancer	Evidence says no, or unsafe	IARC review of more than 1,000 studies found no such effect5
Vitamin D pills prevent the many diseases linked to low vitamin D	Evidence says no, or unsafe	A 2014 review of 172 trials found no effect on disease occurrence; VITAL later saw a borderline drop in autoimmune disease915
Cocoa extract improves thinking skills	Evidence says no, or unsafe	COSMOS-Mind, with 2,262 people, found no benefit, per McGill’s summary2
Getting nutrients from food, not pills, lowers death rates	Promising but early	One large observational study; leftover confounding possible8

Questions readers ask

The bottom line

A correlation is a starting point, not a verdict. The chocolate graph, the coffee reversal, the healthy supplement user and the vitamin D puzzle all show the same thing: two things can travel together for reasons that have nothing to do with one causing the other.

Before you act on a headline, ask about a hidden factor, a backward arrow and plain luck, then look for a trial. That habit protects your health and your wallet, whoever is selling the claim.

Related reading

• Why observational studies got smoking right and hormone therapy wrong
• Simple health answers feel true, so how do you tell which ones are?
• How to weigh healing stories, which move us most when the stakes are highest
• Who paid for the study matters, even when the science is real
• Most health products aren’t scams, but many just haven’t been proven

Sources

1. Husten L. “Chocolate and Nobel Prizes Linked.” CardioExchange, NEJM Journal Watch, October 10, 2012 (summary of Messerli FH, “Chocolate Consumption, Cognitive Function, and Nobel Laureates,” New England Journal of Medicine, 2012). blogs.nejm.org
2. Kellner E. “No, Drinking Milk Will Not Increase Your Chances of Winning a Nobel Prize.” Office for Science and Society, McGill University, accessed October 2026. mcgill.ca
3. Nieburg O. “Eating Chocolate Produces Nobel Prize Winners, Says Study.” FoodNavigator, October 11, 2012. foodnavigator.com
4. Vaz Carneiro A. “Comment on ‘Association of Coffee Drinking With Total and Cause-Specific Mortality'” (summary of Freedman ND et al., New England Journal of Medicine, 2012). Revista Portuguesa de Cardiologia, 2012. revportcardiol.org
5. International Agency for Research on Cancer. “IARC Monographs Evaluate Drinking Coffee, Maté, and Very Hot Beverages.” Press Release No. 244, World Health Organization, June 15, 2016. iarc.who.int
6. International Agency for Research on Cancer. “Debunking the Myths: IARC Monographs Volume 116.” World Health Organization, 2016. iarc.who.int
7. Dickinson A, MacKay D. “Health Habits and Other Characteristics of Dietary Supplement Users: A Review.” Nutrition Journal, 2014. link.springer.com
8. Chen F, Du M, Blumberg JB, et al. “Association Among Dietary Supplement Use, Nutrient Intake, and Mortality Among U.S. Adults: A Cohort Study.” Annals of Internal Medicine, 2019. pubmed.ncbi.nlm.nih.gov
9. Autier P, Boniol M, Pizot C, Mullie P. “Vitamin D Status and Ill Health: A Systematic Review.” The Lancet Diabetes & Endocrinology, 2014 (abstract, University of Strathclyde repository). strathprints.strath.ac.uk
10. Sleight P. “Debate: Subgroup Analyses in Clinical Trials: Fun to Look At, but Don’t Believe Them.” Current Controlled Trials in Cardiovascular Medicine, 2000. link.springer.com
11. Cates C. “The Perils and Pitfalls of Subgroup Analysis.” Testing Treatments Interactive. en.testingtreatments.org
12. The Embryo Project Encyclopedia. “‘The Environment and Disease: Association or Causation?’ (1965), by Austin Bradford Hill.” Arizona State University. embryo.asu.edu
13. Aronson J. “When I Use a Word . . . Austin Bradford Hill’s Viewpoints.” The BMJ Opinion, March 29, 2018. blogs.bmj.com
14. Tufts University. “Nutrients From Food, Not Supplements, Linked to Lower Risks of Death, Cancer.” Tufts Now, April 9, 2019. now.tufts.edu
15. Hahn J, Cook NR, Alexander EK, et al. “Vitamin D and Marine Omega 3 Fatty Acid Supplementation and Incident Autoimmune Disease: VITAL Randomized Controlled Trial.” BMJ, 2022 (copy hosted by Brigham and Women’s Hospital). pubmed.ncbi.nlm.nih.gov